- 3 % discount on prepayment
- Insured shipping from Switzerland
- Sustainable: refurbished instead of new
- Warranty on all cards
- 3 % discount on prepayment
- Insured shipping from Switzerland
- Sustainable: refurbished instead of new
- Warranty on all cards
← Back to overview
Mistral-Small-4-119B-2603
Mistral AI
GermanApache 2.0
Open language model for chat and text. 3 sizes from 119.4B to 675B parameters, partly mixture-of-experts (only a part of the parameters is active per token).
- Mixture-of-experts: 119.4B parameters in total, of which 7.1B active per token. Memory requirement depends on all parameters, speed on the active ones.
Settings
Quality
4 bit (Q4_K_M) is the most common choice: good quality with low memory requirements.
Context length: 8k tokens
4k256k
How much text the model keeps in view at once. More context needs more memory.
Recommendation
1× H100 80 GB
The model needs about 73.3 GB of memory. Hopper, FP8 support.
- Weights
- 71.6 GB
- KV cache
- 0.2 GB
- Speed
- ~234.9 Tok./s
More suitable configurations
| Card | Quantity | Tok./s | Status |
|---|---|---|---|
| Tesla T4 16 GB | 8× | 26.1 | 2-3 weeks delivery time |
| Tesla P100 16 GB | 8× | 32.8 | 2-3 weeks delivery time |
| Tesla P40 24 GB | 4× | 38.7 | 2-3 weeks delivery time |
| Tesla V100 16 GB | 8× | 72.6 | only 1 in stock |
| Tesla V100 32 GB | 4× | 91.3 | On order, delivery time approx. 1 week |
| AMD Instinct MI50 16 GB | 8× | 55.6 | only 2 in stock |
| AMD Instinct MI50 32 GB | 4× | 70.0 | only 2 in stock |
| AMD Instinct MI60 32 GB | 4× | 64.6 | 2-3 weeks delivery time |
| AMD Instinct MI100 32 GB | 4× | 78.2 | 2-3 weeks delivery time |
| A30 24 GB | 4× | 88.3 | 2-3 weeks delivery time |
| RTX A5000 24 GB | 4× | 92.0 | 2-3 weeks delivery time |
| A40 48 GB | 2× | 100.8 | 2-3 weeks delivery time |
